How Cisco UCS affects virtual storage

By Marc Staimer
Tuesday, 14 July, 2009


Cisco Systems Inc. is an acknowledged worldwide leader and renown goliath in the computer networking business. So it's not hard to imagine the anxiety coming from the boardrooms of data center server vendors after Cisco announced a new type of enterprise server system called the Cisco Unified Computing System (UCS). But this new market category raises questions for data storage professionals that we'll try to answer here.

Table of contents:

What market problems does the Unified Computing System solve?
How unique is Cisco's Unified Computing System?
Unified Computing System is new market category, but alternatives exist

What market problems does the Unified Computing System solve?

Cisco maintains that its Unified Computing System solves a series of difficult data center problems while unleashing the potential of virtual servers. Among the market problems Cisco claims UCS addresses are:

  • An inability to get the most out of server virtualisation because of imperfect integration with IP networking, storage networking and storage
  • The headaches of completely separate fabrics for IP networking and storage networking. This increases floor space, rack space, power and cooling, while decreasing IT administrator productivity and infrastructure resource utilisation
  • Limited end-to-end visibility, which reduces IT flexibility and makes it difficult to resolve problems
  • Inefficient energy consumption

But are these issues legitimate, urgent data center storage problems? Ongoing discussions between this author and hundreds of end users and dozens of value-added resellers (VARs) confirm that these are genuine problems when implementing server virtualisation.

How unique is Cisco's Unified Computing System?

Cisco's UCS consists of the following 10 parts.

1. UCS B-Series Blade Servers. The blade servers are based on Intel Corp.'s latest Xeon 5500 (code-named Nehalem). The Xeon 5500 integrates the memory controller and CPU cores, can address more memory and is more energy efficient. However, the Unified Computing System isn't unique in offering Xeon 5500 blades. All of the server vendors have Xeon 5500 server blades or soon will.

There are two things that are unique about the UCS B-Series Blade Servers. They're packaged with considerably more memory, which increases the ratio of potential virtual servers hosted per blade, and a customer can have any type of blade server they want as long as it's a Xeon 5500. In other words, there's extremely limited blade server flexibility with no options for other types of blades such as those from Advanced Micro Devices (AMD) Inc. or even other Intel servers.

2. UCS Converged Network Adapters (CNAs). Converged network adapters allow server TCP/IP network I/O and storage I/O (Fibre Channel over Ethernet [FCoE] and iSCSI) to run concurrently on the same adapters. The UCS CNAs are from QLogic and Emulex. And while there's only one other competitor that has announced optional CNAs at this time (InteliCloud Inc.), it's generally expected that all of the server blade vendors will have it as an option either this year or next as the ANSI standards for Converged Enhanced Ethernet (CEE) are locked down. More on CEE later and whether there are any other proven lower cost alternatives.

Cisco Systems calls its version of CEE switches and CNAs Data Center Ethernet (DCE). But since the standard isn't yet set in stone, it is, by definition, proprietary.

3. UCS 5100 Series Blade Server Chassis. The Unified Computing System blade server chassis is designed for up to eight blade servers and takes up a bit less rack space while somewhat reducing the power and cooling to comparable blade server chassis. Less real estate, power and cooling is always a good thing. But it's the ability to combine up to 40 of the chassis (320 servers) in a single integrated system that adds credence to Cisco's claims of solving data center problems.

4. UCS 2100 Series Fabric Extenders. The fabric extenders are essentially 10 Gigabit Ethernet DCE fan out switches within the server chassis designed to simplify cabling to the blade server chassis. Each fabric extender provides 10 GbE ports per server blade with four 10 GbE CEE ports to the fabric. There can be up to two fabric extenders per chassis, providing a worst-case scenario of 2:1 oversubscription to the chassis.

Chassis fan out switches are far from new or innovative. Fibre Channel, Ethernet and 10 GbE are available on most blade chassis and have been for a while. Even converged network and storage I/O blade chassis switches have been available with InfiniBand (ex. Dell Inc., Hewlett-Packard [HP] Co. and IBM), and more recently with Multi Root IO Virtualisation (based on PCIe switching) for Oracle Corp.-Sun Microsystems Inc. blade chassis.

But whereas InfiniBand is lower latency, higher performance and lower cost than DCE, it's not perceived as mainstream. And while Data Center Ethernet is Cisco Systems' proprietary implementation of CEE, it's based on 10 Gb Ethernet, which is perceived as mainstream and open, and will by definition interoperate with Cisco networks.

The uniqueness here is that Cisco Systems is first to market with converged network and storage I/O on 10 GbE. Converged I/O (regardless of the technology) reduces adapters, cables, switch ports, switches, management, power and cooling.

5. UCS VN-Link (aka Cisco Nexus 1000V Virtual Switch). The VN-Link virtual switch replaces the VMware virtual switch. VN-Link links the virtual and physical networking policies and management. It eliminates the VM firewall that essentially exists to the network administrator and provides the end-to-end visibility that's usually missing. This eliminates the ad hoc scripts that are often written (and not supported) in an attempt to provide some of the same visibility. Its Ethernet-centric design makes it an ideal fit for converged networking and storage I/O.

This component is currently unique to Cisco Systems. However, it can be licensed and implemented on any VMware implementation, including other vendor blade systems.

6. UCS 6100 Series Fabric Interconnects. This is Cisco's DCE interconnect switch. It interconnects the blade chassis to each other, as well as to FCoE storage, FC storage, iSCSI storage or network-attached storage (NAS). Each switch can have 20 or 40 10 GbE ports with up to 20 or 40 switches in a single system. Again, Cisco Systems has a first-to-market advantage with albeit a proprietary version of CEE with gateway capabilities to Fibre Channel.

But there are competitive switches emerging from Brocade/Foundry Networks Inc. Others are expected from Extreme Networks, Force10 Networks Inc. and Juniper Networks Inc., among others.

7. UCS Manager. Cisco manages the entire Unified Computing System virtual data center with one UCS manager platform. The core of the UCS manager is BMC Software Inc.'s "enterprise proven" BladeLogic provisioning of virtual elements software. It also integrates seamlessly with VMware Inc.'s VMware vCenter Server management software to simplify end-to-end control.

This end-to-end management isn't unique. Akorri Inc. BalancePoint, HP OpenView, IBM Tivoli Storage Manager (TSM), Tek-Tools Software Inc. Profiler and Veeam Software Reporter and Monitor products, among others, all have excellent, even outstanding end-to-end server system, network, storage and virtualisation management.

8. OEM software partners with prepackaged integration. This includes the following:

  • VMware vSphere 4 server virtualisation and its vCenter Server for virtual network policies and resources
  • Microsoft Windows Server 2003, Windows Server 2008 Hyper-V and Microsoft SQL Server
  • BMC BladeLogic solutions for provisioning of virtual elements
  • RedHat and Novell SUSE Linux server

9. Storage partners. This includes:

  • EMC Corp.'s Symmetrix V-Max, Clariion and NS Series storage
  • NetApp FAS and V-Series storage

10. Service and support partner. Accenture provides field support services.

Unified Computing System is new market category, but alternatives exist

However, alternatives to Cisco's Unified Computing System do exist. Hewlett-Packard (HP) Co.'s HP BladeSystem Matrix also has a tight integration of servers, virtualisation, management, and network and storage I/O. But unlike the Unified Computing System, the HP BladeSystem Matrix allows for a wide range of server blades, a wealth of storage options and has a complete package of integrated services. InteliCloud's 360 is another option that has an innovative chassis design, a proactive approach to cooling and flexible server blades.

UCS is more than a new Cisco product, it's a new market category that solves noteworthy, urgent data center server virtualisation operational problems. However, UCS isn't the only product set that solves those data center problems. Hewlett-Packard has an excellent alternative with its HP BladeSystem Matrix and InteliCloud has a more technically innovative solution with its 360 product.

Related Articles

How to prepare for the AI future (that isn't here yet)

Something big is indeed coming, but the revolution is not here yet.

Storage strategy in the multicloud era

Data has become the essential raw material in the strategic orientation of business, making data...

Private AI models: redefining data privacy and customisation

Private AI signifies a critical step forward towards a more secure, personalised and efficient...


  • All content Copyright © 2024 Westwick-Farrow Pty Ltd